Re: [PATCH 2/2] arm64: dts: imx8mp: Add support for Data Modul i.MX8M Plus eDM SBC

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On 1/16/23 11:51, Frieder Schrempf wrote:
On 16.01.23 11:39, Marek Vasut wrote:
On 1/11/23 08:38, Ahmad Fatoum wrote:
Hello Marek,

Hi,

[...]

Could you share your get_maintainers.pl invocation? I'd like to
adjust the reviewer entry in MAINTAINERS, so such patches get
into our kernel@xxxxxxxxxxxxxx inbox as well.

Plain get_maintainer -f path/to/dts

Some more comments below.

+    pmic: pmic@25 {
+        compatible = "nxp,pca9450c";
+        reg = <0x25>;
+        pinctrl-names = "default";
+        pinctrl-0 = <&pinctrl_pmic>;
+        interrupt-parent = <&gpio1>;
+        interrupts = <3 IRQ_TYPE_LEVEL_LOW>;
+        sd-vsel-gpios = <&gpio1 4 GPIO_ACTIVE_HIGH>;

Here you assume GPIO1_IO04 is muxed as GPIO.

+    pinctrl_usdhc2: usdhc2-grp {
+        fsl,pins = <
+            MX8MP_IOMUXC_GPIO1_IO04__USDHC2_VSELECT        0xc1

Here you mux it for USDHC2_VSELECT though. Is this intended?

That's a good question, other DTs do the same thing, I suspect the
sd-vsel-gpios could be dropped,

The sd-vsel-gpios is only needed when the hardware has an arbitrary GPIO
connected to the SD_VSEL signal of the PCA9450. IMHO, if SD_VSEL is
connected to the VSELECT signal it would be better to drop sd-vsel-gpios.

Current practice does not match this statement however, notice how they are all gpio 1 4 :

$ git grep sd-vsel-gpios
Documentation/devicetree/bindings/regulator/nxp,pca9450-regulator.yaml: sd-vsel-gpios: arch/arm64/boot/dts/freescale/imx8mm-verdin.dtsi: sd-vsel-gpios = <&gpio1 4 GPIO_ACTIVE_HIGH>; arch/arm64/boot/dts/freescale/imx8mp-data-modul-edm-sbc.dts: sd-vsel-gpios = <&gpio1 4 GPIO_ACTIVE_HIGH>; arch/arm64/boot/dts/freescale/imx8mp-dhcom-som.dtsi: sd-vsel-gpios = <&gpio1 4 GPIO_ACTIVE_HIGH>; arch/arm64/boot/dts/freescale/imx8mp-verdin.dtsi: sd-vsel-gpios = <&gpio1 4 GPIO_ACTIVE_HIGH>;

but as long as it is not outright
harmful, it can be used for backward compatibility to support less
complete OSes which may not handle the eSDHC VSELECT bit , so I figured
it is good to keep both options.

Would be interesting what OSes you have in mind? The eSDHC
driver/hardware should always handle the VSELECT signal, no?

Not necessarily, but the ones I am aware of (U-Boot, Linux) do handle VSELECT.



[Index of Archives]     [Device Tree Compilter]     [Device Tree Spec]     [Linux Driver Backports]     [Video for Linux]     [Linux USB Devel]     [Linux PCI Devel]     [Linux Audio Users]     [Linux Kernel]     [Linux SCSI]     [XFree86]     [Yosemite Backpacking]


  Powered by Linux